{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"A practical deep-dive into building reusable helper functions, zero-shot and few-shot classification, SQL generation, clinical information extraction, multilingual translation, RAG pipelines, and domain-specific chatbot architecture — all using the OpenAI API."}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"📑 Table of Contents"}]},{"type":"orderedList","attrs":{"start":1,"type":null},"content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#1-setup--environment","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Setup & Environment"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#2-role-based-messaging--the-get_completion-helper","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Role-Based Messaging & the "},{"type":"text","marks":[{"type":"code"}],"text":"get_completion"},{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#2-role-based-messaging--the-get_completion-helper","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":" Helper"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#3-output-formatting-in-jupyter","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Output Formatting in Jupyter"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#4-zero-shot-vs-few-shot-classification","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Zero-Shot vs Few-Shot Classification"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#5-sentiment-analysis-pipeline","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Sentiment Analysis Pipeline"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#6-temperature-in-practice","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Temperature in Practice"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#7-evaluation--accuracy-metrics","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Evaluation & Accuracy Metrics"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#8-llm-prompted-machine-learning-pipeline","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"LLM-Prompted Machine Learning Pipeline"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#9-deep-learning-extension--mlp-with-keras","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Deep Learning Extension — MLP with Keras"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#10-sql-joins--subqueries","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"SQL Joins & Subqueries"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#11-prompt-engineering-for-sql-generation","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Prompt Engineering for SQL Generation"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#12-building-domain-specific-chatbots","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Building Domain-Specific Chatbots"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#13-rag-pipeline--step-by-step","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"RAG Pipeline — Step by Step"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#14-information-extraction-from-clinical-notes","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Information Extraction from Clinical Notes"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#15-multilingual-translation","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Multilingual Translation"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#16-default-parameters--model-selection","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Default Parameters & Model Selection"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"link","attrs":{"href":"https://claude.ai/chat/e0586dbe-16f4-49af-8d70-ec3ad622f1fd#17-troubleshooting-checklist","target":"_blank","rel":"noopener noreferrer nofollow","class":"text-blue-600 underline"}}],"text":"Troubleshooting Checklist"}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"1. Setup & Environment"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Installing the package"}]},{"type":"codeBlock","attrs":{"language":"bash"},"content":[{"type":"text","text":"pip install openai\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Configuring the API key securely"}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"API key"},{"type":"text","text":" — a secret token that authenticates your requests to the OpenAI service. Never hard-code it in source files."}]}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"from getpass import getpass\nimport os\nimport openai\n\n# Prompt interactively — input is hidden\nopenai_api_key = getpass(\"Enter your OpenAI API key: \")\n\n# Option A: set as an environment variable\nos.environ[\"OPENAI_API_KEY\"] = openai_api_key\n\n# Option B: assign directly on the openai module\nopenai.api_key = openai_api_key\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Importing display utilities for Jupyter"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"from IPython.display import display, Markdown, HTML\n"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"These utilities let you render model output as formatted Markdown or HTML inside notebook cells — essential for readable tables, bullet lists, and headings."}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"2. Role-Based Messaging & the "},{"type":"text","marks":[{"type":"code"}],"text":"get_completion"},{"type":"text","text":" Helper"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Message roles"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Role"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Purpose"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"system"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Sets the behaviour or persona of the assistant"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"user"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Supplies the actual query or instruction"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"assistant"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Represents the model's previous response (used in multi-turn dialogs)"}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Core "},{"type":"text","marks":[{"type":"code"}],"text":"get_completion"},{"type":"text","text":" helper"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"import openai\n\ndef get_completion(prompt, model=\"gpt-4o-mini\", temperature=0):\n \"\"\"\n Send a single user prompt and return the model's text response.\n\n Parameters\n ----------\n prompt : str The instruction or question for the model.\n model : str Model identifier (default: gpt-4o-mini).\n temperature : float Randomness control (0 = deterministic, 1 = creative).\n \"\"\"\n messages = [{\"role\": \"user\", \"content\": prompt}]\n\n response = openai.ChatCompletion.create(\n model=model,\n messages=messages,\n temperature=temperature\n )\n\n return response.choices[0].message[\"content\"]\n"}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Default parameter:"},{"type":"text","text":" When "},{"type":"text","marks":[{"type":"code"}],"text":"model"},{"type":"text","text":" is not supplied by the caller, "},{"type":"text","marks":[{"type":"code"}],"text":"\"gpt-4o-mini\""},{"type":"text","text":" is used automatically. Passing "},{"type":"text","marks":[{"type":"code"}],"text":"model=\"gpt-4o\""},{"type":"text","text":" overrides the default, enabling easy experimentation with speed vs. quality vs. cost."}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Extended helper — with system prompt"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"def get_completion_with_system(system_prompt, user_prompt,\n model=\"gpt-4o-mini\", temperature=0):\n messages = [\n {\"role\": \"system\", \"content\": system_prompt},\n {\"role\": \"user\", \"content\": user_prompt}\n ]\n response = openai.ChatCompletion.create(\n model=model,\n messages=messages,\n temperature=temperature\n )\n return response.choices[0].message[\"content\"]\n"}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"3. Output Formatting in Jupyter"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"The model often returns Markdown-formatted text. Rendering it properly improves readability."}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Markdown rendering"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"from IPython.display import display, Markdown\n\nresult = get_completion(\"List five benefits of RAG in three bullet points.\")\ndisplay(Markdown(result))\n"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Output (rendered):"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Reduces hallucination by grounding answers in retrieved documents"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Enables up-to-date answers without model retraining"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Keeps proprietary data out of the model weights"}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"HTML rendering"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"from IPython.display import display, HTML\n\nhtml_table = \"\"\"\n\n \n \n

Review	Sentiment
Great product!	Positive

\n\"\"\"\ndisplay(HTML(html_table))\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"When to use which"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Format"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Use When"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"display(Markdown(...))"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Model returns bullet lists, headings, bold text, code blocks"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"display(HTML(...))"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"You need richer layout — colours, custom tables, embedded links"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"print(...)"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Plain debugging output; no formatting needed"}]}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"4. Zero-Shot vs Few-Shot Classification"}]},{"type":"customImage","attrs":{"src":"https://67dc5f63b8b6a6f0ea39b99a.storage.fermion.app/public-cached/markdown-editor-file/file-unique-id/jhpapkovmcnfjwueebtgnoa8nnwhcecijfcttagncc91a_s-_sgnagpcvf6pyo8d5a5bblc88g3bm5s8tx9oozhlnwdfmna2fr5q","alt":null,"width":600,"height":327}},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Definitions"}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Zero-shot learning"},{"type":"text","text":" — performing a task without providing any examples; the model infers the task from the instruction alone."}]}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Few-shot learning"},{"type":"text","text":" — supplying a handful of labelled input-output pairs so the model can mimic the demonstrated pattern."}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Side-by-side comparison"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Aspect"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Zero-Shot"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Few-Shot"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Examples provided"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"None"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"3–10 labelled examples embedded in the prompt"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Prompt length"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Short"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Longer (examples increase token count)"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Typical accuracy"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Good baseline"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Usually higher — examples guide the model"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"When to use"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Quick prototyping, no labelled data available"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Slightly higher accuracy when representative cases exist"}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Zero-shot prompt template"}]},{"type":"codeBlock","attrs":{"language":"javascript"},"content":[{"type":"text","text":"Act as a product review analyst.\nGiven the following review, display the overall sentiment as ONLY one of:\nPositive, Negative, Neutral.\n\nReview: {review_text}\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Few-shot prompt template"}]},{"type":"codeBlock","attrs":{"language":"javascript"},"content":[{"type":"text","text":"Act as a product review analyst.\n\nReview: \"The new phone works flawlessly and the battery lasts all day.\" → Positive 👍\nReview: \"The laptop crashes constantly and the screen flickers.\" → Negative 👎\n\nReview: {review_text}\n"}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"The model uses the demonstrated pattern to classify the new review. Adding emoji (👍 👎) to the examples trains the model to include them in its output — useful for visual dashboards."}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"5. Sentiment Analysis Pipeline"}]},{"type":"customImage","attrs":{"src":"https://67dc5f63b8b6a6f0ea39b99a.storage.fermion.app/public-cached/markdown-editor-file/file-unique-id/g4urippenpukdebaazrswucvb3ywwd6mtqtvkjgu_0fm_rb0wf0zlah3c4o1rcuxmsab59juhzexdlj_jgqrya_8lbsb_698bzag","alt":null,"width":600,"height":327}},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Sample reviews"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"reviews = [\n \"The Bluetooth speaker has clear sound, good bass, and is waterproof.\",\n \"The kitchen blender is noisy, jams frequently, and arrived late.\",\n \"The laptop has an excellent display but the keyboard feels cheap.\",\n \"Delivery was fast and packaging was secure. Very happy with the purchase!\"\n]\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Batch classification loop"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"import openai\n\ndef classify_sentiment(review: str) -> str:\n prompt = f\"\"\"Act as a product review analyst.\nGiven the following review, display the overall sentiment as ONLY one of:\nPositive, Negative, Neutral.\n\nReview: {review}\"\"\"\n return get_completion(prompt, temperature=0)\n\n# Process all reviews\nresponses = []\nfor review in reviews:\n sentiment = classify_sentiment(review)\n responses.append(sentiment)\n print(f\"Review: {review[:50]}... → {sentiment}\")\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Using "},{"type":"text","marks":[{"type":"code"}],"text":"zip"},{"type":"text","text":" for parallel lists"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"When you have a list of reviews and a list of true labels, "},{"type":"text","marks":[{"type":"code"}],"text":"zip"},{"type":"text","text":" pairs them for simultaneous processing:"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"reviews = [\"Great product!\", \"Terrible, broke after a day.\", \"Average, nothing special.\"]\nlabels = [\"Positive\", \"Negative\", \"Neutral\"]\n\nfor review, true_label in zip(reviews, labels):\n predicted = classify_sentiment(review)\n match = \"✅\" if predicted.strip() == true_label else \"❌\"\n print(f\"{match} Predicted: {predicted:10s} | True: {true_label}\")\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Displaying results as a Markdown table"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"from IPython.display import display, Markdown\n\ntable_prompt = f\"\"\"\nAct as a product review analyst.\nClassify the sentiment of each review in the list below.\nReturn the results as a Markdown table with columns: | Review | Sentiment |\n\nReviews:\n{chr(10).join(f'- {r}' for r in reviews)}\n\"\"\"\n\nresult = get_completion(table_prompt, temperature=0)\ndisplay(Markdown(result))\n"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Sample output (rendered):"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Review"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Sentiment"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"The Bluetooth speaker has clear sound..."}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Positive"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"The kitchen blender is noisy and jams..."}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Negative"}]}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"6. Temperature in Practice"}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Temperature"},{"type":"text","text":" controls the randomness of the model's output."}]}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Temperature"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Behaviour"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"0"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Deterministic — same answer every time; ideal for classification, factual queries"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"0.7 – 1.0"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Moderately creative — varied phrasing, nuanced sentiment"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"> 1.0"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Highly creative / imaginative — use for brainstorming or poetry"}]}]}]}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"# Precise, consistent classification\nsentiment = get_completion(prompt, temperature=0)\n\n# More creative, varied response\nstory_opening = get_completion(\n \"Write an opening line for a mystery novel.\",\n temperature=1.0\n)\n"}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"For production classification pipelines, always use "},{"type":"text","marks":[{"type":"code"}],"text":"temperature=0"},{"type":"text","text":" to ensure reproducible, deterministic output."}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"7. Evaluation & Accuracy Metrics"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"After classifying many reviews, measure model quality against a manually labelled validation set."}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"from sklearn.metrics import accuracy_score, classification_report\n\ntrue_labels = [\"Positive\", \"Negative\", \"Neutral\", \"Positive\"]\npredicted_labels = [classify_sentiment(r).strip() for r in reviews]\n\nprint(\"Accuracy:\", accuracy_score(true_labels, predicted_labels))\nprint(classification_report(true_labels, predicted_labels))\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Evaluation workflow"}]},{"type":"orderedList","attrs":{"start":1,"type":null},"content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Validate a subset manually"},{"type":"text","text":" — verify 50–100 predictions by hand"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Iterate on the prompt"},{"type":"text","text":" — add few-shot examples, refine wording, adjust temperature"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Track metrics"},{"type":"text","text":" — log accuracy, precision, recall, F1 for each prompt version"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Document results"},{"type":"text","text":" — record which prompt version was used for reproducibility"}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"8. LLM-Prompted Machine Learning Pipeline"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"One of the most powerful use cases for LLMs: "},{"type":"text","marks":[{"type":"bold"}],"text":"generating complete, runnable ML code"},{"type":"text","text":" from a structured prompt."}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Prompt for a Titanic classification task"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"titanic_prompt = \"\"\"\nAct as an expert data-science assistant.\nGenerate Python code that:\n\n1. Loads the Titanic CSV from a public URL.\n2. Performs preprocessing:\n - Drop irrelevant columns: PassengerId, Name, Ticket, Cabin\n - Encode the Sex column (male=0, female=1)\n - Fill missing Age values with the median\n - Fill missing Embarked values with the mode\n3. Conducts EDA with:\n - Summary statistics (df.describe())\n - Visualisations: correlation heatmap, age distribution histogram,\n survival rate by gender bar chart\n4. Splits data into training (75%) and test (25%) sets.\n5. Trains three models:\n - Logistic Regression\n - Decision Tree\n - Random Forest\n6. Evaluates each model using:\n - Confusion matrix\n - Classification report\n - Accuracy score\n7. Prints a comparison table and recommends the top-performing model.\n\"\"\"\n\ncode = get_completion(titanic_prompt, temperature=0)\nprint(code)\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"The 7-step ML pipeline (structure the LLM generates)"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Step"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Task"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"1"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Pre-processing — missing values, encoding, scaling"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"2"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"EDA — descriptive statistics, visualisations (heatmaps)"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"3"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Data visualisation — age distribution, survival by gender"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"4"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Train-test split"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"5"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Model building — Logistic Regression, Decision Tree, Random Forest"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"6"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Evaluation — confusion matrix, classification report, accuracy"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"7"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Conclusion — compare models, select best for production"}]}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"9. Deep Learning Extension — MLP with Keras"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Extend the pipeline with a neural network by adjusting the prompt:"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"keras_prompt = \"\"\"\nAct as an expert deep-learning practitioner.\nWrite Python code using TensorFlow and Keras to build a Multi-Layer Perceptron\nfor the Titanic dataset.\n\nInclude:\n- Data loading and preprocessing (same as before).\n- Model architecture:\n Input → Dense(64, activation='relu')\n → Dense(32, activation='relu')\n → Output Dense(1, activation='sigmoid')\n- Compile with:\n optimizer='adam'\n loss='binary_crossentropy'\n metrics=['accuracy']\n- Train the model for 50 epochs with a 20% validation split.\n- Plot training vs. validation loss and accuracy over epochs.\n- Evaluate on the test set and print final accuracy.\n\"\"\"\n\nkeras_code = get_completion(keras_prompt, temperature=0)\nprint(keras_code)\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Generated architecture (illustrative)"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"from tensorflow import keras\nfrom tensorflow.keras import layers\n\nmodel = keras.Sequential([\n layers.Dense(64, activation='relu', input_shape=(X_train.shape[1],)),\n layers.Dense(32, activation='relu'),\n layers.Dense(1, activation='sigmoid')\n])\n\nmodel.compile(\n optimizer='adam',\n loss='binary_crossentropy',\n metrics=['accuracy']\n)\n\nhistory = model.fit(\n X_train, y_train,\n epochs=50,\n validation_split=0.2,\n batch_size=32,\n verbose=1\n)\n"}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"10. SQL Joins & Subqueries"}]},{"type":"paragraph","attrs":{"textAlign":null}},{"type":"customImage","attrs":{"src":"https://learnsql.com/blog/learn-and-practice-sql-joins/2.png","alt":"SQL Joins","width":600,"height":338}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Join"},{"type":"text","text":" — an operation that combines rows from two or more tables based on a related column."}]}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Subquery"},{"type":"text","text":" — a query nested inside another query, used to supply a condition that depends on the result of the inner query."}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Common join types"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Join Type"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Description"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Syntax Example"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"INNER JOIN"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Returns rows with matching values in "},{"type":"text","marks":[{"type":"bold"}],"text":"both"},{"type":"text","text":" tables"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"FROM employee e INNER JOIN department d ON e.dept_id = d.dept_id"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"LEFT JOIN"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"All rows from left table; unmatched right rows → NULL"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"LEFT JOIN salary s ON e.emp_id = s.emp_id"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"RIGHT JOIN"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"All rows from right table; unmatched left rows → NULL"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"RIGHT JOIN ..."}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"FULL JOIN"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Rows when there is a match in "},{"type":"text","marks":[{"type":"bold"}],"text":"either"},{"type":"text","text":" table"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"FULL JOIN ..."}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"CROSS JOIN"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Cartesian product — every combination of rows"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"CROSS JOIN dept d"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"NATURAL JOIN"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Joins on columns with the same name, implicitly"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"NATURAL JOIN employee"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"UNION"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Stacks result sets; removes duplicates"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"SELECT ... FROM t1 UNION SELECT ... FROM t2"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"UNION ALL"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Stacks result sets; keeps duplicates"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"... UNION ALL ..."}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Subqueries — why they matter"}]},{"type":"codeBlock","attrs":{"language":"sql"},"content":[{"type":"text","text":"-- Find employees whose salary is greater than the average salary of their department\nSELECT emp_name, salary\nFROM employee e\nJOIN salary s ON e.emp_id = s.emp_id\nWHERE s.salary > (\n SELECT AVG(s2.salary)\n FROM salary s2\n JOIN employee e2 ON s2.emp_id = e2.emp_id\n WHERE e2.dept_id = e.dept_id\n);\n"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Subqueries can appear in "},{"type":"text","marks":[{"type":"code"}],"text":"SELECT"},{"type":"text","text":", "},{"type":"text","marks":[{"type":"code"}],"text":"WHERE"},{"type":"text","text":", or "},{"type":"text","marks":[{"type":"code"}],"text":"FROM"},{"type":"text","text":" clauses, enabling condition-inside-condition logic."}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Practical example — Top 5 highest-paid in Data Science"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Schema:"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Table"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Columns"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"department"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"dept_id, dept_name"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"employee"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"emp_id, emp_name, dept_id"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"salary"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"emp_id, salary"}]}]}]}]},{"type":"codeBlock","attrs":{"language":"sql"},"content":[{"type":"text","text":"SELECT\n e.emp_name,\n d.dept_name,\n s.salary\nFROM employee e\nJOIN department d ON e.dept_id = d.dept_id\nJOIN salary s ON e.emp_id = s.emp_id\nWHERE d.dept_name = 'Data Science'\nORDER BY s.salary DESC\nLIMIT 5;\n"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Key clauses:"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"e"},{"type":"text","text":", "},{"type":"text","marks":[{"type":"code"}],"text":"d"},{"type":"text","text":", "},{"type":"text","marks":[{"type":"code"}],"text":"s"},{"type":"text","text":" are "},{"type":"text","marks":[{"type":"bold"}],"text":"aliases"},{"type":"text","text":" for readability — shorter than writing the full table name"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"ORDER BY ... DESC"},{"type":"text","text":" sorts from highest to lowest"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"LIMIT 5"},{"type":"text","text":" returns only the top 5 rows"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"SQL is "},{"type":"text","marks":[{"type":"bold"}],"text":"case-insensitive"},{"type":"text","text":" — "},{"type":"text","marks":[{"type":"code"}],"text":"SELECT"},{"type":"text","text":" and "},{"type":"text","marks":[{"type":"code"}],"text":"select"},{"type":"text","text":" behave identically"}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"11. Prompt Engineering for SQL Generation"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"You can use an LLM to write SQL queries on demand by providing the schema and task in the prompt."}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Generic SQL generation prompt"}]},{"type":"codeBlock","attrs":{"language":"javascript"},"content":[{"type":"text","text":"Act as an expert in generating SQL code.\n\nUnderstand the following schema:\n- department(dept_id, dept_name)\n- employee(emp_id, emp_name, dept_id)\n- salary(emp_id, salary)\n\nWrite a query that returns the top 5 employees with the highest salary\nin the {department_name} department, showing employee name, department name,\nand salary, ordered from highest to lowest.\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Key elements of an effective SQL prompt"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Element"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Purpose"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Role specification"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"\"Act as an expert in generating SQL code.\""}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Schema description"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"List all tables with their column names"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Task description"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"What exactly the query must return"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Output requirements"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Columns to include, ordering, row limit"}]}]}]}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"def generate_sql(schema: str, task: str) -> str:\n prompt = f\"\"\"Act as an expert SQL developer.\n\nSchema:\n{schema}\n\nTask: {task}\n\nReturn ONLY the SQL query — no explanation, no markdown fences.\"\"\"\n return get_completion(prompt, temperature=0)\n\nschema = \"\"\"\n- department(dept_id, dept_name)\n- employee(emp_id, emp_name, dept_id)\n- salary(emp_id, salary)\n\"\"\"\n\nquery = generate_sql(schema,\n \"Top 5 highest-paid employees in the Engineering department, \"\n \"showing name, dept, and salary sorted descending.\")\nprint(query)\n"}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"12. Building Domain-Specific Chatbots"}]},{"type":"customImage","attrs":{"src":"https://67dc5f63b8b6a6f0ea39b99a.storage.fermion.app/public-cached/markdown-editor-file/file-unique-id/gmcvujijplhywsur8pzecvgqfug1qp08oovdyavl8ah-fmjm11chelgnul92v94uyc-1wji8dvuoupdyav1hqxyghdmsihcsjc6e","alt":null,"width":600,"height":327}},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Architecture overview"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Layer"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Purpose"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Typical Technologies"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Backend"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Executes prompts, calls LLM APIs, manages retrieval pipelines"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Python, FastAPI, OpenAI SDK"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Frontend"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Captures user input, renders LLM output"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"React, Streamlit, VSCode extensions"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Data Store"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Persists custom knowledge (company documents)"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Vector databases (FAISS, Pinecone, Chroma)"}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Why self-host a chatbot?"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Privacy"},{"type":"text","text":" — avoids third-party data leakage"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Control"},{"type":"text","text":" — integrate proprietary documents via RAG"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Customisation"},{"type":"text","text":" — define exact scope through the system prompt"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Cost management"},{"type":"text","text":" — fine-tune model choice and token usage"}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Key implementation steps"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"from fastapi import FastAPI\nfrom openai import OpenAI\nimport os\n\napp = FastAPI()\nclient = OpenAI(api_key=os.getenv(\"OPENAI_API_KEY\"))\n\nSYSTEM_PROMPT = \"\"\"You are an SQL assistant specialised in database queries.\n- Only answer SQL-related questions.\n- If a question is off-topic, say: 'I can only assist with SQL queries.'\n- Always provide runnable, well-commented SQL.\"\"\"\n\n@app.post(\"/chat\")\ndef chat(user_message: str):\n response = client.chat.completions.create(\n model=\"gpt-4o-mini\",\n messages=[\n {\"role\": \"system\", \"content\": SYSTEM_PROMPT},\n {\"role\": \"user\", \"content\": user_message}\n ],\n temperature=0\n )\n return {\"reply\": response.choices[0].message.content}\n"}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"13. RAG Pipeline — Step by Step"}]},{"type":"customImage","attrs":{"src":"https://67dc5f63b8b6a6f0ea39b99a.storage.fermion.app/public-cached/markdown-editor-file/file-unique-id/8btqp1ndlwqerepxjae-6qjd0v2vp5y98xkxcckrmqgvslb79spoztcsdvwtlh8immbrti9exgw_8q107uehejxeuft932brzcba","alt":null,"width":600,"height":327}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"RAG (Retrieval-Augmented Generation)"},{"type":"text","text":" — a pipeline that retrieves relevant information from a knowledge base and augments the LLM prompt with that context before generation."}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Minimal RAG flow"}]},{"type":"codeBlock","attrs":{"language":"javascript"},"content":[{"type":"text","text":"User Query → Embed → Retrieve (vector DB) → Augment Prompt → LLM → Answer → Evaluation → Output\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"8-step detailed pipeline"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Step"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Name"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Description"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Technologies"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"1"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Document Loader"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Ingest raw files — PDFs, CSVs, web pages"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"LangChain loaders, PyPDF2"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"2"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Chunking"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Split documents into manageable pieces (~500-token chunks)"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"RecursiveCharacterTextSplitter"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"3"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Embedding"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Convert each chunk to a dense vector"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"text-embedding-ada-002"},{"type":"text","text":", HuggingFace"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"4"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Vector Store / Index"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Store vectors in a searchable database"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"FAISS, Pinecone, Chroma, Elasticsearch"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"5"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Retrieval"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Compute cosine similarity; return top-k relevant chunks"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Vector DB similarity search"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"6"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Augmentation"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Append retrieved chunks to the user prompt"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Prompt templating"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"7"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"LLM Generation"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Feed the augmented prompt to the LLM"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"OpenAI, Anthropic, Gemini"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"8"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Evaluation & Control"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Hallucination check, guardrails, feedback loop"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"RAGAS, TruLens"}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Minimal LangChain RAG implementation"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"from langchain_community.document_loaders import PyPDFLoader\nfrom langchain.text_splitter import RecursiveCharacterTextSplitter\nfrom langchain_openai import OpenAIEmbeddings\nfrom langchain_community.vectorstores import FAISS\nfrom langchain_openai import ChatOpenAI\nfrom langchain.chains import RetrievalQA\n\n# 1. Load document\nloader = PyPDFLoader(\"company_policy.pdf\")\ndocs = loader.load()\n\n# 2. Chunk\nsplitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)\nchunks = splitter.split_documents(docs)\n\n# 3 + 4. Embed and store\nvectorstore = FAISS.from_documents(chunks, OpenAIEmbeddings())\n\n# 5. Retriever\nretriever = vectorstore.as_retriever(search_kwargs={\"k\": 3})\n\n# 6 + 7. RAG chain\nqa = RetrievalQA.from_chain_type(\n llm=ChatOpenAI(model=\"gpt-4o-mini\", temperature=0),\n retriever=retriever\n)\n\n# Query\nanswer = qa.run(\"What is the refund policy for hardware products?\")\nprint(answer)\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Risk controls in RAG"}]},{"type":"bulletList","content":[{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Content filtering"},{"type":"text","text":" — block harmful or off-topic output"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Provenance tags"},{"type":"text","text":" — cite the source chunk in every answer"}]}]},{"type":"listItem","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Feedback loop"},{"type":"text","text":" — log user corrections to improve retrieval quality"}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"14. Information Extraction from Clinical Notes"}]},{"type":"customImage","attrs":{"src":"https://67dc5f63b8b6a6f0ea39b99a.storage.fermion.app/public-cached/markdown-editor-file/file-unique-id/vyavgp6rdqneucjlnguqnkcc5e7xhwwbui9p2_osb4iuqx6y2czxy1vsmgivl9lq5u7mbqmipvjvhapd99jadvzwxalbf3qxo0vp","alt":null,"width":600,"height":327}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Information extraction"},{"type":"text","text":" — identifying and structuring specific entities (e.g., symptoms, diagnoses) from unstructured clinical text."}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Prompt for symptom extraction"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"def extract_symptoms(clinical_note: str) -> str:\n prompt = f\"\"\"Act as an expert in analyzing clinical doctor notes.\n\nExtract all symptoms mentioned in the text below.\n\nFor each symptom:\n1. Indicate Presence: Present / Absent / Denies\n2. Assign Priority: High / Medium / Low\n3. Provide a brief Explanation for the priority.\n4. Do NOT invent symptoms — only list those explicitly mentioned.\n\nExpand all medical acronyms (e.g., CAD → Coronary Artery Disease).\n\nReturn results as a Markdown table with columns:\n| Symptom | Presence | Priority | Explanation |\n\nThen include an Appendix table:\n| Acronym | Expanded Form | Meaning |\n\nClinical note:\n{clinical_note}\"\"\"\n\n return get_completion(prompt, temperature=0)\n\n# Example note\nnote = \"\"\"\nPatient presents with swelling of the tongue and throat. Denies chest pain.\nNo nausea. PMH includes HTN and DM2. CAD diagnosed 3 years ago.\n\"\"\"\n\nresult = extract_symptoms(note)\ndisplay(Markdown(result))\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Expected output format (Symptom Table)"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Symptom"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Presence"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Priority"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Explanation"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Swelling of the tongue"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Present"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"High"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Airway obstruction risk"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Swelling of the throat"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Present"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"High"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Respiratory compromise"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Chest pain"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Absent"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Low"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Explicitly denied by patient"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Nausea"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Absent"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Low"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Not mentioned in note"}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Appendix — Acronym Table"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Acronym"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Expanded Form"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Meaning"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"HTN"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Hypertension"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"High blood pressure"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"DM2"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Diabetes Mellitus Type 2"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Chronic metabolic disorder"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"CAD"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Coronary Artery Disease"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Cardiovascular condition"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"PMH"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Past Medical History"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Prior health conditions"}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Probability categories"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Category"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Description"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"High"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Strong evidence symptom is present; model assigns high confidence"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Medium"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Moderate evidence; confidence lower than High but still notable"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Low"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Weak or absent evidence"}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Why RAG improves clinical extraction"}]},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Clinical notes contain domain-specific jargon and abbreviations. Loading a corpus of medical literature into the vector store gives the LLM "},{"type":"text","marks":[{"type":"bold"}],"text":"contextual grounding"},{"type":"text","text":", reducing hallucinations and improving acronym expansion accuracy."}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"15. Multilingual Translation"}]},{"type":"customImage","attrs":{"src":"https://67dc5f63b8b6a6f0ea39b99a.storage.fermion.app/public-cached/markdown-editor-file/file-unique-id/dz0pclc9wd-ejbbdirobvpkoeenwiehwl9dvrwhmkerdx0dlrdfh0gspxzzqjkqqrvnolrwomkocfhq9mhi1gmzbvqkuo-f6rvsi","alt":null,"width":600,"height":327}},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Universal translation prompt"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"def translate_text(text: str, target_languages: list) -> str:\n lang_list = \", \".join(target_languages)\n prompt = f\"\"\"You are an expert translator.\nTranslate the following English text into these languages: {lang_list}\n\nFor each language, label the output clearly with the language name.\nIf a language name is not recognised, return a clear error message.\n\nText: {text}\"\"\"\n return get_completion(prompt, temperature=0)\n\n# Example\nenglish_text = \"Artificial intelligence is transforming every industry.\"\n\ntranslations = translate_text(english_text, [\n \"Hindi\", \"Bengali\", \"Tamil\", \"Telugu\", \"French\", \"Japanese\", \"Arabic\"\n])\ndisplay(Markdown(translations))\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Supported scripts & languages"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Language"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Script"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Hindi"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Devanagari"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Bengali"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Bengali"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Tamil"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Tamil"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Telugu"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Telugu"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Marathi"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Devanagari"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Gujarati"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Gujarati"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Malayalam"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Malayalam"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Kannada"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Kannada"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Punjabi"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Gurmukhi"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"French"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Latin"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Chinese"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Han characters"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Spanish"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Latin"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Japanese"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Kana / Kanji"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Arabic"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Arabic (right-to-left)"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Urdu"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Arabic script (right-to-left)"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Sanskrit"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Devanagari"}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Practical tips for translation tasks"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Tip"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Guidance"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Single-language requests"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Faster; avoid overloading with many languages per call"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Chunking long texts"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Split to ≤ 8,191-token chunks before sending"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Temperature"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Use "},{"type":"text","marks":[{"type":"code"}],"text":"temperature=0"},{"type":"text","text":" for deterministic, formal translations"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Batch processing"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Loop over language list; collect responses in a dict"}]}]}]}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"# Batch: one call per language for precision\ndef batch_translate(text: str, languages: list) -> dict:\n results = {}\n for lang in languages:\n prompt = f\"Translate to {lang}: '{text}'\"\n results[lang] = get_completion(prompt, temperature=0)\n return results\n\ntranslations = batch_translate(\n \"Data science combines statistics, programming, and domain expertise.\",\n [\"Hindi\", \"Tamil\", \"French\"]\n)\n\nfor lang, translation in translations.items():\n print(f\"{lang}: {translation}\")\n"}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"16. Default Parameters & Model Selection"}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"bold"}],"text":"Default parameter"},{"type":"text","text":" — a value assigned to a function argument that is used when the caller does not provide an explicit value."}]}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"# \"gpt-4o-mini\" is the default — used when no model is specified\ndef get_completion(prompt, model=\"gpt-4o-mini\", temperature=0):\n ...\n\n# Default (fast, cheap)\nresult = get_completion(\"Explain transformers.\")\n\n# Override to a more capable model\nresult = get_completion(\"Explain transformers.\", model=\"gpt-4o\")\n\n# Override to an older model for legacy compatibility\nresult = get_completion(\"Explain transformers.\", model=\"gpt-3.5-turbo\")\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Model selection guide"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Model"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Speed"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Cost"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Quality"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Best For"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"gpt-3.5-turbo"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"⚡⚡⚡"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"💰"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"★★★"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Simple tasks, high-volume batches"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"gpt-4o-mini"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"⚡⚡"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"💰💰"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"★★★★"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Balanced — best default choice"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"gpt-4o"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"⚡"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"💰💰💰"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"★★★★★"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Complex reasoning, code generation"}]}]}]}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"17. Troubleshooting Checklist"}]},{"type":"table","content":[{"type":"tableRow","content":[{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Symptom"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Cause"}]}]},{"type":"tableHeader","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Fix"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"ModuleNotFoundError: openai"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Package not installed"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"pip install openai"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"AuthenticationError"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"API key missing or wrong"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Set "},{"type":"text","marks":[{"type":"code"}],"text":"OPENAI_API_KEY"},{"type":"text","text":" environment variable"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"RateLimitError"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Too many requests / insufficient credits"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Check billing; add credits; implement retry with back-off"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"InvalidRequestError: max_tokens"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"max_tokens"},{"type":"text","text":" exceeds model limit"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Reduce "},{"type":"text","marks":[{"type":"code"}],"text":"max_tokens"},{"type":"text","text":" or switch to a higher-context model"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"Token limit exceeded"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Input prompt is too long"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Split text into ≤ 8,191-token chunks"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"code"}],"text":"unknown parameter: quality"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Typo in parameter value"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Use "},{"type":"text","marks":[{"type":"code"}],"text":"\"standard\""},{"type":"text","text":" or "},{"type":"text","marks":[{"type":"code"}],"text":"\"hd\""},{"type":"text","text":" — check spelling"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Access denied on "},{"type":"text","marks":[{"type":"code"}],"text":"pip install"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Missing admin rights"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Run with administrator privileges or use a virtual environment"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Kernel reset in Colab"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Free-tier timeout"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Store key in environment file; re-run setup cells"}]}]}]},{"type":"tableRow","content":[{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Slow responses"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Large batches on CPU"}]}]},{"type":"tableCell","attrs":{"colspan":1,"rowspan":1,"colwidth":null},"content":[{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","text":"Use Google Colab GPU or reduce batch size"}]}]}]}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Runtime & kernel management"}]},{"type":"codeBlock","attrs":{"language":"python"},"content":[{"type":"text","text":"# Kernel reset in Colab forces re-entry of the API key\n# Pattern to minimise disruption: define all helpers in one cell at the top\n\ndef setup():\n \"\"\"Run once after every kernel restart.\"\"\"\n from getpass import getpass\n import openai\n openai.api_key = getpass(\"API key: \")\n print(\"✅ Ready\")\n\nsetup()\n"}]},{"type":"horizontalRule"},{"type":"heading","attrs":{"textAlign":null,"level":2},"content":[{"type":"text","text":"🎯 Quick Reference — Prompt Templates"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Sentiment classification (zero-shot)"}]},{"type":"codeBlock","attrs":{"language":"javascript"},"content":[{"type":"text","text":"Act as a product review analyst.\nGiven the following review, display the overall sentiment as ONLY one of:\nPositive, Negative, Neutral.\n\nReview: {review}\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"SQL generation"}]},{"type":"codeBlock","attrs":{"language":"javascript"},"content":[{"type":"text","text":"Act as an expert SQL developer.\nSchema: {schema}\nTask: {task}\nReturn ONLY the SQL query.\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Clinical symptom extraction"}]},{"type":"codeBlock","attrs":{"language":"javascript"},"content":[{"type":"text","text":"Act as an expert in analyzing clinical doctor notes.\nExtract all symptoms. For each: Presence, Priority, Explanation.\nExpand acronyms. Return as Markdown tables.\nNote: {note}\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"Multilingual translation"}]},{"type":"codeBlock","attrs":{"language":"javascript"},"content":[{"type":"text","text":"You are an expert translator.\nTranslate the following into {languages}: {text}\nLabel each translation with the language name.\n"}]},{"type":"heading","attrs":{"textAlign":null,"level":3},"content":[{"type":"text","text":"ML code generation"}]},{"type":"codeBlock","attrs":{"language":"javascript"},"content":[{"type":"text","text":"Act as an expert data-science assistant.\nGenerate Python code that:\n1. Loads {dataset}\n2. Preprocesses: {preprocessing_steps}\n3. Trains: {models}\n4. Evaluates with: confusion matrix, classification report, accuracy.\n5. Recommends the best model.\n"}]},{"type":"horizontalRule"},{"type":"paragraph","attrs":{"textAlign":null},"content":[{"type":"text","marks":[{"type":"italic"}],"text":"Applied prompt engineering transforms the OpenAI API from a raw capability into a production-ready tool — whether you're classifying thousands of reviews, querying databases in natural language, extracting insights from medical records, or translating content across 20 languages."}]}]}